{
  "id": 9253794,
  "title": "Alibaba targets 10 trillion parameters with next generation Qwen 4 model",
  "url": "https://urgent.news/2026/09/23/alibaba-targets-10-trillion-parameters-with-next-generation-qwen-4",
  "topic": "business",
  "section": "Business",
  "published": "2026-09-23T03:03:48.000Z",
  "source": {
    "name": "Vietnam Investment Review",
    "slug": "vietnam-investment-review",
    "url": "https://vir.com.vn/alibaba-targets-10-trillion-parameters-with-next-generation-qwen-4-model-161322.html"
  },
  "original_language": "en",
  "account": "Alibaba, a leading technology company, has announced significant advancements in its AI strategy and capabilities. At the Apsara Conference, Alibaba unveiled a roadmap for its full-stack AI approach, aiming to unleash the full potential of AI at scale for its customers. Joe Tsai, Chairman of Alibaba Group, emphasized that machine thinking has only reached 3% of human thinking capacity, with the potential to scale to 1,000x human capacity in the future. Eddie Wu, CEO of Alibaba Group, added that machines are becoming the primary force behind thinking, and groundbreaking products of the Machine Intelligence era are yet to arrive.\n\nTo support these objectives, Alibaba unveiled several new innovations and updates. The company will deploy AI chips and a purpose-built agentic cloud, along with an AI agent platform for mobile devices. A key focus is on the development of Qwen foundation models and multimodal models, with a target of 5 to 10 trillion parameters in the Qwen 4.5 and Qwen 5 series. Alibaba has also made progress in Recursive Self-Improvement (RSI) through fully automated runs, improving model performance significantly.\n\nIn the realm of speech and audio processing, Alibaba introduced Qwen3.8-LiveTranslate, a simultaneous interpretation model that enhances translation quality and reduces latency. Additionally, Qwen-Audio-3.1-TTS-Next, an advanced audio generation model, is designed for professional creative applications such as audiobooks, film, and television. The company also unveiled Qwen-Audio-3.1-ASR, Qwen-Audio-3.1-TTS, and Qwen-Audio-3.1-Realtime for improved voice AI capabilities.\n\nFor visual content creation, Alibaba showcased Qwen-Image 3.1, an image-generation model optimized for creative design and e-commerce marketing. The model offers advanced features such as transparent-background generation and versatile image-editing capabilities. To support these advancements, Alibaba will complete the deployment of its AI infrastructure with a total data center capacity of over 20GW by 2032, enabling the industry to meet the rising demand for AI.",
  "summary": "Technology conglomerate Alibaba introduced custom artificial intelligence chips and outlined cloud infrastructure developments at its Apsara Conference to support models scaling to 10 trillion parameters",
  "key_points": [
    "Alibaba aims to develop Qwen 4 model with 5 to 10 trillion parameters.",
    "Qwen3.8-LiveTranslate introduced for simultaneous interpretation.",
    "Alibaba plans to deploy AI infrastructure with over 20GW capacity by 2032."
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}